diff --git a/English/ai-stack/agent-base/agent-runtime/openclaw/deploy-and-manage-openclaw.md b/English/ai-stack/agent-base/agent-runtime/openclaw/deploy-and-manage-openclaw.md
index 4ecf7e88..c1625e52 100644
--- a/English/ai-stack/agent-base/agent-runtime/openclaw/deploy-and-manage-openclaw.md
+++ b/English/ai-stack/agent-base/agent-runtime/openclaw/deploy-and-manage-openclaw.md
@@ -73,6 +73,10 @@ Once all fields are filled, click **"Start Setup"** to begin provisioning.
The **"Setting Up Your Workspace"** screen displays a loading spinner while the system automatically prepares your environment. Once complete, you receive a **Gateway Token** and the **OpenClaw web admin URL** to log in and start using immediately.
+{% hint style="warning" %}
+**Save the Gateway Token the moment it appears.** This token is **shown only once** on the success screen — afterwards it is hidden and cannot be retrieved. Copy it somewhere safe (a password manager) before you leave the page. If you have already lost it, contact the GreenNode team for help.
+{% endhint %}
+
#### Step 3: Deploy Success
@@ -83,7 +87,7 @@ Once provisioning is complete, the Deploy Success screen shows your instance det
| ------------------------ | ------------------------------------------------------------ |
| **Instance Name** | The name of your created instance (e.g. `openclaw/username`) |
| **Status** | 🟢 Active |
-| **Gateway Token** | Token used to log in to the OpenClaw web admin |
+| **Gateway Token** | Token used to log in to the OpenClaw web admin — **shown only once**, copy and save it immediately |
| **OpenClaw Gateway URL** | Link to your OpenClaw web admin |
| **Created At** | Timestamp |
diff --git a/English/ai-stack/agent-base/faq.md b/English/ai-stack/agent-base/faq.md
index b1ece9e8..22f82a77 100644
--- a/English/ai-stack/agent-base/faq.md
+++ b/English/ai-stack/agent-base/faq.md
@@ -158,6 +158,37 @@ curl -s -X PATCH "https://agentbase.api.vngcloud.vn/runtime/agent-runtimes/$RUNT
***
+**Q: My agent needs somewhere to store data and files. Does AgentBase have Storage?**
+
+AgentBase has **no** storage service of its own. The runtime container filesystem is **ephemeral** — anything written inside the container is lost when a replica restarts, scales, or when you deploy a new version. Don't treat it as durable storage.
+
+For durable storage, use **S3 storage via vStorage**. Your agent uploads files to S3 and retrieves them when needed, using any S3-compatible client (e.g. `boto3`):
+
+```python
+import boto3
+
+s3 = boto3.client(
+ "s3",
+ endpoint_url="",
+ aws_access_key_id="",
+ aws_secret_access_key="",
+)
+
+# Upload a file the agent produced
+s3.upload_file("/tmp/report.pdf", "my-bucket", "reports/report.pdf")
+
+# Retrieve it later
+s3.download_file("my-bucket", "reports/report.pdf", "/tmp/report.pdf")
+```
+
+Create your bucket and access keys in the [vStorage Console](https://vstorage.console.greennode.ai/overview).
+
+{% hint style="info" %}
+Store the vStorage access key as an [Access Control](access-control/) auth configuration rather than hardcoding it in your image — the agent retrieves the credential at runtime via its agent identity.
+{% endhint %}
+
+***
+
## Memory
**Q: Is my conversation data private to my organization?**
diff --git a/English/ai-stack/model-as-a-service/available-models.md b/English/ai-stack/model-as-a-service/available-models.md
index 9f5a8b30..35106e19 100644
--- a/English/ai-stack/model-as-a-service/available-models.md
+++ b/English/ai-stack/model-as-a-service/available-models.md
@@ -6,6 +6,23 @@ The MaaS API provides full-accuracy access to model families such as Anthropic,
The model catalog below has been updated under the transition starting **August 3, 2026**, covering models self-hosted by GreenNode and third-party models under official contracts. See the timeline details in [Release Notes](../release-notes.md) and the billing rules in [Pricing](pricing.md).
{% endhint %}
+### Rate limit
+
+Applies by default to **every account** using the MaaS API:
+
+| Limit | Value |
+| --- | --- |
+| Per minute | **10 RPM** — 10 requests/minute |
+| Per day | **14,400 requests/day** |
+
+The limit is counted **per account** and shared across all models — it is not a separate per-model quota. In the tables below, a `–` in the **Rate Limit** column means that model has no limit of its own and falls under this default.
+
+{% hint style="info" %}
+Need more than 10 RPM? [Contact GreenNode](https://helpdesk.greennode.ai/portal/en/newticket) to be considered for the **whitelist**. Whitelisting is reviewed case by case.
+{% endhint %}
+
+***
+
### Model list
The **Provider Type** column tells you where your request is processed:
@@ -65,6 +82,3 @@ Modalities describe the model's general capability per its provider's documentat
| ------------------------ | ------------- | ------------------------------- | ---------- | --------------- |
| Cohere Rerank v4.0 pro | Third-party | Text → Ranked List | – | August 20, 2026 |
-Note:
-
-* You can create [ticket](https://helpdesk.greennode.ai/portal/vi/newticket) for request new model
diff --git a/Vietnamese/ai-stack/agent-base/agent-runtime/openclaw/trien-khai-openclaw-1-click.md b/Vietnamese/ai-stack/agent-base/agent-runtime/openclaw/trien-khai-openclaw-1-click.md
index 297104be..0a9925f4 100644
--- a/Vietnamese/ai-stack/agent-base/agent-runtime/openclaw/trien-khai-openclaw-1-click.md
+++ b/Vietnamese/ai-stack/agent-base/agent-runtime/openclaw/trien-khai-openclaw-1-click.md
@@ -73,6 +73,10 @@ Sau khi điền đầy đủ thông tin, nhấn **"Bắt đầu thiết lập"**
Màn hình **"Setting Up Your Workspace"** hiển thị icon xoay trong khi hệ thống tự động chuẩn bị môi trường. Sau khi hoàn tất, bạn nhận được **Gateway Token** và **URL trang web admin OpenClaw** để đăng nhập và sử dụng ngay.
+{% hint style="warning" %}
+**Lưu Gateway Token ngay khi nó hiện ra.** Token này **chỉ hiển thị một lần** ở màn hình tạo thành công — sau đó sẽ bị ẩn và không xem lại được. Copy và lưu vào nơi an toàn (password manager) trước khi rời khỏi trang. Nếu bạn đã mất token, hãy liên hệ đội ngũ GreenNode để được hỗ trợ.
+{% endhint %}
+
#### Bước 3: Deploy Success
@@ -83,7 +87,7 @@ Sau khi provisioning hoàn tất, màn hình Deploy Success hiển thị thông
| ------------------------ | ------------------------------------------------ |
| **Instance Name** | Tên instance đã tạo (ví dụ: `openclaw/username`) |
| **Trạng thái** | 🟢 Active |
-| **Gateway Token** | Token đăng nhập vào trang web admin OpenClaw |
+| **Gateway Token** | Token đăng nhập vào trang web admin OpenClaw — **chỉ hiển thị một lần**, hãy copy và lưu lại ngay |
| **OpenClaw Gateway URL** | Link truy cập trang web admin OpenClaw |
| **Thời gian tạo** | Timestamp |
diff --git a/Vietnamese/ai-stack/agent-base/faq.md b/Vietnamese/ai-stack/agent-base/faq.md
index eb4c6ee3..9cd2f98b 100644
--- a/Vietnamese/ai-stack/agent-base/faq.md
+++ b/Vietnamese/ai-stack/agent-base/faq.md
@@ -158,6 +158,37 @@ curl -s -X PATCH "https://agentbase.api.vngcloud.vn/runtime/agent-runtimes/$RUNT
***
+**H: Agent của tôi cần nơi lưu trữ data và file. AgentBase có Storage không?**
+
+AgentBase **không có** dịch vụ storage riêng. Filesystem của container runtime là **ephemeral** — mọi file ghi trong container sẽ mất khi replica restart, scale, hoặc khi bạn deploy phiên bản mới. Đừng dùng nó làm nơi lưu trữ lâu dài.
+
+Để lưu trữ bền vững, dùng **S3 storage thông qua vStorage**. Agent của bạn upload file lên S3 và retrieve lại khi cần, bằng bất kỳ S3-compatible client nào (ví dụ `boto3`):
+
+```python
+import boto3
+
+s3 = boto3.client(
+ "s3",
+ endpoint_url="",
+ aws_access_key_id="",
+ aws_secret_access_key="",
+)
+
+# Upload file agent vừa tạo
+s3.upload_file("/tmp/report.pdf", "my-bucket", "reports/report.pdf")
+
+# Retrieve lại khi cần
+s3.download_file("my-bucket", "reports/report.pdf", "/tmp/report.pdf")
+```
+
+Khởi tạo bucket và lấy access key tại [vStorage Console](https://vstorage.console.greennode.ai/overview).
+
+{% hint style="info" %}
+Lưu access key của vStorage bằng cấu hình xác thực của [Access Control](access-control/) thay vì hardcode trong image — agent truy xuất credential lúc runtime theo agent identity.
+{% endhint %}
+
+***
+
## Memory
**H: Dữ liệu hội thoại của tôi có riêng tư cho tổ chức không?**
diff --git a/Vietnamese/ai-stack/model-as-a-service/cac-model-duoc-cung-cap.md b/Vietnamese/ai-stack/model-as-a-service/cac-model-duoc-cung-cap.md
index 3b79fcb8..aae01834 100644
--- a/Vietnamese/ai-stack/model-as-a-service/cac-model-duoc-cung-cap.md
+++ b/Vietnamese/ai-stack/model-as-a-service/cac-model-duoc-cung-cap.md
@@ -6,6 +6,23 @@ MaaS API cung cấp quyền truy cập với độ chính xác đầy đủ vào
Danh mục model bên dưới đã được cập nhật theo lộ trình chuyển đổi bắt đầu **03/08/2026**, gồm model do GreenNode self-host và model third-party đã ký hợp đồng chính thức. Xem chi tiết mốc thời gian tại [Release Notes](../release-notes.md) và nguyên tắc tính phí tại [Cách tính phí](cach-tinh-phi.md).
{% endhint %}
+### Rate limit
+
+Mặc định áp dụng cho **mọi tài khoản** dùng MaaS API:
+
+| Giới hạn | Giá trị |
+| --- | --- |
+| Theo phút | **10 RPM** — 10 request/phút |
+| Theo ngày | **14.400 request/ngày** |
+
+Giới hạn được tính trên **tài khoản** và dùng chung cho mọi model — không phải hạn mức riêng của từng model. Trong các bảng bên dưới, cột **Rate Limit** hiển thị `–` nghĩa là model đó không có giới hạn riêng và áp dụng mức mặc định này.
+
+{% hint style="info" %}
+Cần mức cao hơn 10 RPM? Hãy [liên hệ GreenNode](https://helpdesk.greennode.ai/portal/vi/newticket) để được xem xét đưa vào **whitelist**. Whitelist được xét theo từng trường hợp cụ thể.
+{% endhint %}
+
+***
+
### Danh sách mô hình
Cột **Loại Provider** cho biết request của bạn được xử lý ở đâu:
@@ -65,7 +82,3 @@ Modalities mô tả khả năng chung của model theo tài liệu của nhà cu
| ------------------------ | ------------- | ------------------------------- | ---------- | ---------- |
| Cohere Rerank v4.0 pro | Third-party | Text → Ranked List | – | 20/08/2026 |
-Lưu ý:
-
-* Bạn có thể tạo [ticket](https://helpdesk.greennode.ai/portal/vi/newticket) để yêu cầu model mong muốn
-